Together AI Raises $800M as Enterprise Demand for Open-Source Cloud Triples

The San Francisco startup Together AI has closed an $800 million Series C funding round at an $8.3 billion post-money valuation, cementing its position as the dominant “neocloud” for open-source AI workloads. The round was led by Saudi Aramco’s venture arm, Aramco Ventures, with participation from Vista Equity Partners, General Catalyst, Emergence Capital, Nvidia, March Capital, Pegatron, and SentinelOne’s S Ventures.

The raise more than doubles Together AI’s previous valuation of $3.3 billion, set during its $305 million Series B roughly 16 months ago. But the number that caught Wall Street’s attention wasn’t the valuation: it was the company’s annual bookings, which have surpassed $1.15 billion — a milestone that puts Together squarely in revenue territory most AI startups only dream about.

The Open-Source Tipping Point

Together AI’s business model is deceptively simple but the timing has proven prescient: rent Nvidia GPU clusters and purpose-built AI infrastructure to enterprises that want to run open-source models at costs far below what proprietary frontier providers charge. Rather than paying the premium per-token rates commanded by OpenAI or Anthropic, customers can deploy models like DeepSeek, Meta’s Llama family, MiniMax, and Kimi at a fraction of the price, with full control over data and deployment architecture.

The company says usage of open-source models across its platform has tripled over the past year. That surge tracks with a broader industry reckoning visible across enterprise AI procurement: as AI becomes infrastructure rather than a novelty, CFOs are scrutinizing token bills the same way they once dissected cloud compute invoices. Tesla’s announcement this week that it would cap employee AI spending at $200 per week — a dramatic reversal from its earlier push for aggressive adoption — is the most visible symptom of a cost-control wave sweeping corporate America.

“The frontier model market is changing fast, and enterprises don’t want to be locked into one vendor’s pricing,” said CEO Vipul Ved Prakash in remarks accompanying the announcement. “Open-source models give them the flexibility to benchmark, switch, and optimize — and Together gives them the infrastructure to do it reliably at scale.”

Who’s Building on Together

The customer list reads like a directory of the AI developer ecosystem’s most ambitious players. Cursor, the AI-powered code editor that has become the default productivity tool for software engineers, runs inference on Together’s platform. So does Cognition, maker of the autonomous coding agent Devin, and Decagon, which builds AI-native customer support systems. The common thread: these are fast-moving, AI-first companies for whom inference costs are a first-order business variable, not a rounding error in the cloud bill.

Together’s infrastructure pitch extends beyond price. The company has secured commitments for over 500 megawatts of dedicated compute capacity — a pipeline that will be capitalized independently by strategic investors, keeping the capital-intensive hardware off Together’s balance sheet while ensuring supply can meet demand surges. At current GPU power densities, 500 megawatts represents a substantial fraction of the global high-performance AI compute pool outside the hyperscaler giants.

Aramco Ventures’ leadership of the round is strategically revealing. Saudi Arabia’s state oil giant has been aggressive in its AI investment strategy, seeking to diversify both its own operations and the kingdom’s economy through technology partnerships. Anchoring the dominant open-source AI infrastructure play fits neatly into that national diversification thesis, while also giving Aramco preferential access to cost-efficient AI infrastructure for its own digital transformation initiatives.

Neocloud vs. Hyperscaler

Together’s rise coincides with a structural reconfiguration of the AI compute market. The traditional hyperscalers — AWS, Azure, Google Cloud — have been slow to fully embrace open-source model hosting at scale, often steering customers toward their own proprietary AI services and ecosystems. That gap created the opening for neoclouds: purpose-built GPU infrastructure providers that treat open-source models as first-class citizens rather than afterthoughts.

Together isn’t alone in this space. CoreWeave, Lambda Labs, and Fireworks AI are all competing for the same enterprise dollars. But Together has cultivated an edge in the software layer: its platform offers fine-tuning pipelines, model customization workflows, and API compatibility layers that allow developers to migrate from proprietary APIs without rewriting their application code. For an enterprise already invested in a ChatGPT-compatible integration, switching to Together is closer to a configuration change than a re-architecture.

The Nvidia participation in this round carries particular strategic weight. For Jensen Huang’s company, the neocloud wave represents a massive GPU demand multiplier: every enterprise that shifts from cloud-hosted proprietary inference to independently managed open-source deployments needs more hardware, not less. Nvidia backing Together is effectively a bet that the expansion of open-source AI will expand the total addressable market for its silicon, even as competition among model providers intensifies.

The Cost Reckoning and What Comes Next

The funding arrives at a critical inflection point for the AI industry’s economics. OpenAI’s GPT-5.6 family, unveiled last week with three tiers (Sol, Terra, and Luna) explicitly optimized for different cost-performance points, signals that even the proprietary leaders are scrambling to address enterprise cost sensitivity. Google’s Gemini team is navigating its own pricing pressure. The implicit message from Together’s $1.15 billion bookings number: the open-source alternative is no longer a cost-conscious startup workaround. It is a mainstream enterprise option at serious scale.

For Vipul Ved Prakash and co-founders Percy Liang and Ce Zhang — all veterans of Stanford’s Center for Research on Foundation Models who built CRFM before founding Together in 2022 — the funding represents validation of a founding thesis: that open access to frontier AI capabilities, not lock-in to any single provider, is the architecture that will determine who controls the AI stack long-term.

With $800 million in fresh capital, 500 megawatts of compute on the way, and annual bookings north of $1 billion, Together is no longer just making the case for open-source AI. It’s building the financial and physical infrastructure to enforce it.